A non-transitory computer-readable medium having stored thereon computer-readable instructions executable to estimate quantile values for fraud assessments in a fraud detection system, the computer-readable instructions executable to perform quantile value estimation operations which comprises:
- determining a first time period associated with one or more first data values that are designated for updating a quantile value profile usable to perform the fraud assessments, wherein the quantile value profile is associated with an entity;
- capturing, over the first time period, the one or more first data values for the quantile value profile, wherein the quantile value profile is further associated with a second time period and comprises one of (a) real values for one or more second data values for the entity over the second time period or (b) a first plurality of quantile marker values calculated from the one or more second data values;
- at an end of the first time period, accessing the quantile value profile for the entity;
- determining a first number of the one or more first data values over the first time period;
- and
- based on the first number of the one or more first data values and the one of the real values or the first plurality of quantile marker values in the quantile value profile, performing one of a first merge operation, a second merge operation, or a third merge operation on the one or more first data values captured over the first time period with the one of the real values or the first plurality of quantile marker values in the quantile value profile, wherein the performing the second merge operation and the third merge operation includes:
- computing intersections of a parabolic equation between heights and positions of the one or more first data values, the one of the real values, or a combination thereof, and
- determining coefficients of the parabolic equation based on the intersections, wherein an output of the parabolic equation comprises different quantile value markers that reduce the number of one of the one or more first data values or the one of the real values, and wherein the coefficients for a parabola of the parabolic equation intersects different ones of the first plurality of quantile marker values and a second plurality of quantile marker values associated with the quantile value profile based on the intersections;
- after the second time period, performing a profile update for the quantile value profile by:
- determining that a data storage size of the quantile value profile requires a reduction that reduces the quantile value profile to a minimum set of quantile value markers associated with the identified parabolic equation having the determined coefficients, and
- reducing the data storage size of the quantile value profile using the third merge operation;
- and
- performing the fraud assessments using the quantile value profile having the profile update, wherein the minimum set of quantile value markers in the quantile value profile are used to replace past data values processed by an intelligent fraud detection engine for the fraud assessments.